Self vs. Proxy Reporting in the Consumer Expenditure Survey Nancy - PowerPoint PPT Presentation

Self vs. Proxy Reporting in the Consumer Expenditure Survey Nancy A. Mathiowetz University of Wisconsin-Milwaukee 1

Framing the discussion  Nature of the Survey Instrument Frames the Research Questions  Consumer Expenditure Interview Survey  Retrospective Recall  How to minimize response error related to long term recall?  Consumer Expenditure Diary  Completion  How to maximize participation by all members of the CU?  With analytic unit (CU) in mind 2

Note…… We know little with respect to the quality of both self and proxy-based reports for the CE, so research aimed at reducing bias associated with proxy- based reports should not ignore the onerous task facing the self-reporter. 3

Literature…….  Moore: little support for the notion that self response generally better than quality of proxy reports  Exception: Turner’s study of crime  Alwin: Lower reliability for proxy reports  education, occupation, hours work  CE Focus  Intra-Household Communication Study  Self proxy agreement: 63% of the time  Self response: more purchases and higher dollar amounts  JPSM Teen-Parent Study  24 hour period: 85% agreement rate  Mean underreporting dollar amount: $6.75 4

CEI: Cognitive Processing Perspective  Knowledge/Encoding  To what extent does the respondent/recorder have knowledge of the expenditure?  Retrieval  To what extent can the respondent/recorder retrieve the expenditure?  Reporting  To what extent is the respondent/recorder willing to report the expenditure? 5

Focus on improving proxy reporting……  Understand under what conditions proxy reports are more consistent with self  Acquisition of the knowledge, richness of the encoded material  Relationship between self and proxy  Salient vs. mundane purchases; routine or rare occurrences  Focus on question wording to improve retrieval strategies  Episodic vs. Estimation strategies 6

..or focus on improving overall quality?  Expand the use of records  Usual and expected expenditures  Utility bills, mortgages/rent  Receipts  Take advantage of longitudinal nature of CEI  Redesign of information book?  Inter-interview outreach  IVR “tagging” periodically throughout the 3 month period  Incorporate Technology  Barcode readers  Smartphone Applications 7

Diary…  Multiple respondents per HH  Grootaert (1986)  Impact on personal item expenditures  Not necessary to get complete participation to be effective  Arbitron Radio Diary  BLS field study  Increase in the number of expenditure items  Increased dollar value of items  Lower RR; increased number of trips per complete 8

Different problem, similar solution  Expand the use of records  Grocery store receipts, restaurant, convenience stores  Incorporate Technology  Barcode readers  Smartphone Applications 9

Solutions demand full understanding of the problem  CEI: how effective is the bounding interview?  To what extent do respondents understand that the use of records would reduce the burden?  CEI: what retrieval strategies do respondents use?  Example: CEI as source of expenditures for gasoline  CED: what do we know about the communication of the diary to others in the CU?  What reminders exist to help encourage reporting?  To what extent do interviewers stress the need for full participation?  CED: are there privacy issues within CU that suppresses reporting? 10

Cost:Error Tradeoff  Presumptive Nature of Tradeoff Questions  Poorer quality data for proxy  Compared to benchmarks, CE underestimates expenditures. But is this a proxy problem? What do we know about CE quality for self reports?  Lower Response Rates  Nonreponse bias, not rates  Full participation may not be the key to improvement  Higher Costs  Only if implementing the same approach to field operations 11

Burden: Redefine the task  Examine source of expenditure for CPI and focus efforts within each instrument  Reduce Redundancies  “all other products, services, and expenses”  Divide and conquer: focus diary on “personal” expenditures/those for which diary provides input for the CPI and the CEI on household-level  CED: Food, personal care products & services, prescription and non-prescription drugs, housekeeping supplies, clothing  Grootaert findings 12

Before making changes in the design, consider……  Conducting a series of ethnographic and/or observational studies to gather further information about the response process for self and proxy reports.  Designing studies so as to assess measurement error for both self and proxy reports.  Example: partner with Nielson and use their Homescan data as a validation source.  Expanding the 2006 experimental study of the individual CED.  Reduce redundancies between CU level CED and individual diaries.  Include non-response follow-up study to address the tradeoff between increased participation and changes in response rates 13

With respect to design……  Redesign CED so as to capture receipts rather than requiring recording of information by hand.  Reduce the burden by pre-identifying common purchases that can be checked off rather than written in by hand.  Experimenting with more aggressive requests for record keeping that involves all CU members during the first CEI interview.  Support and encourage record keeping through the use of outreach, including but not limited to postcards, IVR, email, and incentives. 14

Technology……  Test the feasibility of incorporating technology in the CED  If technology feasibility study is positive, consider altering the design of the CED  Longer reporting period.  Tradeoff between reduced costs of enrollment vs. increased length of time as a diarist. Examine diary fatigue  CEI and CED within the same CU  Experiment with the use of technology for the CEI 15